ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.198 by root, Thu Oct 23 06:30:48 2008 UTC vs.
Revision 1.246 by root, Thu Jul 2 12:08:55 2009 UTC

…		…
9	=head2 EXAMPLE PROGRAM	9	=head2 EXAMPLE PROGRAM
10		10
11	// a single header file is required	11	// a single header file is required
12	#include <ev.h>	12	#include <ev.h>
13		13
		14	#include <stdio.h> // for puts
		15
14	// every watcher type has its own typedef'd struct	16	// every watcher type has its own typedef'd struct
15	// with the name ev_<type>	17	// with the name ev_TYPE
16	ev_io stdin_watcher;	18	ev_io stdin_watcher;
17	ev_timer timeout_watcher;	19	ev_timer timeout_watcher;
18		20
19	// all watcher callbacks have a similar signature	21	// all watcher callbacks have a similar signature
20	// this callback is called when data is readable on stdin	22	// this callback is called when data is readable on stdin
…		…
41		43
42	int	44	int
43	main (void)	45	main (void)
44	{	46	{
45	// use the default event loop unless you have special needs	47	// use the default event loop unless you have special needs
46	ev_loop *loop = ev_default_loop (0);	48	struct ev_loop *loop = ev_default_loop (0);
47		49
48	// initialise an io watcher, then start it	50	// initialise an io watcher, then start it
49	// this one will watch for stdin to become readable	51	// this one will watch for stdin to become readable
50	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);	52	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
51	ev_io_start (loop, &stdin_watcher);	53	ev_io_start (loop, &stdin_watcher);
…		…
60		62
61	// unloop was called, so exit	63	// unloop was called, so exit
62	return 0;	64	return 0;
63	}	65	}
64		66
65	=head1 DESCRIPTION	67	=head1 ABOUT THIS DOCUMENT
		68
		69	This document documents the libev software package.
66		70
67	The newest version of this document is also available as an html-formatted	71	The newest version of this document is also available as an html-formatted
68	web page you might find easier to navigate when reading it for the first	72	web page you might find easier to navigate when reading it for the first
69	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.	73	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		74
		75	While this document tries to be as complete as possible in documenting
		76	libev, its usage and the rationale behind its design, it is not a tutorial
		77	on event-based programming, nor will it introduce event-based programming
		78	with libev.
		79
		80	Familarity with event based programming techniques in general is assumed
		81	throughout this document.
		82
		83	=head1 ABOUT LIBEV
70		84
71	Libev is an event loop: you register interest in certain events (such as a	85	Libev is an event loop: you register interest in certain events (such as a
72	file descriptor being readable or a timeout occurring), and it will manage	86	file descriptor being readable or a timeout occurring), and it will manage
73	these event sources and provide your program with events.	87	these event sources and provide your program with events.
74		88
…		…
108	name C<loop> (which is always of type C<ev_loop *>) will not have	122	name C<loop> (which is always of type C<ev_loop *>) will not have
109	this argument.	123	this argument.
110		124
111	=head2 TIME REPRESENTATION	125	=head2 TIME REPRESENTATION
112		126
113	Libev represents time as a single floating point number, representing the	127	Libev represents time as a single floating point number, representing
114	(fractional) number of seconds since the (POSIX) epoch (somewhere near	128	the (fractional) number of seconds since the (POSIX) epoch (somewhere
115	the beginning of 1970, details are complicated, don't ask). This type is	129	near the beginning of 1970, details are complicated, don't ask). This
116	called C<ev_tstamp>, which is what you should use too. It usually aliases	130	type is called C<ev_tstamp>, which is what you should use too. It usually
117	to the C<double> type in C, and when you need to do any calculations on	131	aliases to the C<double> type in C. When you need to do any calculations
118	it, you should treat it as some floating point value. Unlike the name	132	on it, you should treat it as some floating point value. Unlike the name
119	component C<stamp> might indicate, it is also used for time differences	133	component C<stamp> might indicate, it is also used for time differences
120	throughout libev.	134	throughout libev.
121		135
122	=head1 ERROR HANDLING	136	=head1 ERROR HANDLING
123		137
…		…
276		290
277	=back	291	=back
278		292
279	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	293	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP
280		294
281	An event loop is described by a C<ev_loop *>. The library knows two	295	An event loop is described by a C<struct ev_loop *> (the C<struct>
282	types of such loops, the I<default> loop, which supports signals and child	296	is I<not> optional in this case, as there is also an C<ev_loop>
283	events, and dynamically created loops which do not.	297	I<function>).
		298
		299	The library knows two types of such loops, the I<default> loop, which
		300	supports signals and child events, and dynamically created loops which do
		301	not.
284		302
285	=over 4	303	=over 4
286		304
287	=item struct ev_loop *ev_default_loop (unsigned int flags)	305	=item struct ev_loop *ev_default_loop (unsigned int flags)
288		306
…		…
294	If you don't know what event loop to use, use the one returned from this	312	If you don't know what event loop to use, use the one returned from this
295	function.	313	function.
296		314
297	Note that this function is I<not> thread-safe, so if you want to use it	315	Note that this function is I<not> thread-safe, so if you want to use it
298	from multiple threads, you have to lock (note also that this is unlikely,	316	from multiple threads, you have to lock (note also that this is unlikely,
299	as loops cannot bes hared easily between threads anyway).	317	as loops cannot be shared easily between threads anyway).
300		318
301	The default loop is the only loop that can handle C<ev_signal> and	319	The default loop is the only loop that can handle C<ev_signal> and
302	C<ev_child> watchers, and to do this, it always registers a handler	320	C<ev_child> watchers, and to do this, it always registers a handler
303	for C<SIGCHLD>. If this is a problem for your application you can either	321	for C<SIGCHLD>. If this is a problem for your application you can either
304	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you	322	create a dynamic loop with C<ev_loop_new> that doesn't do that, or you
…		…
380	=item C<EVBACKEND_EPOLL> (value 4, Linux)	398	=item C<EVBACKEND_EPOLL> (value 4, Linux)
381		399
382	For few fds, this backend is a bit little slower than poll and select,	400	For few fds, this backend is a bit little slower than poll and select,
383	but it scales phenomenally better. While poll and select usually scale	401	but it scales phenomenally better. While poll and select usually scale
384	like O(total_fds) where n is the total number of fds (or the highest fd),	402	like O(total_fds) where n is the total number of fds (or the highest fd),
385	epoll scales either O(1) or O(active_fds). The epoll design has a number	403	epoll scales either O(1) or O(active_fds).
386	of shortcomings, such as silently dropping events in some hard-to-detect	404
387	cases and requiring a system call per fd change, no fork support and bad	405	The epoll mechanism deserves honorable mention as the most misdesigned
388	support for dup.	406	of the more advanced event mechanisms: mere annoyances include silently
		407	dropping file descriptors, requiring a system call per change per file
		408	descriptor (and unnecessary guessing of parameters), problems with dup and
		409	so on. The biggest issue is fork races, however - if a program forks then
		410	I<both> parent and child process have to recreate the epoll set, which can
		411	take considerable time (one syscall per file descriptor) and is of course
		412	hard to detect.
		413
		414	Epoll is also notoriously buggy - embedding epoll fds I<should> work, but
		415	of course I<doesn't>, and epoll just loves to report events for totally
		416	I<different> file descriptors (even already closed ones, so one cannot
		417	even remove them from the set) than registered in the set (especially
		418	on SMP systems). Libev tries to counter these spurious notifications by
		419	employing an additional generation counter and comparing that against the
		420	events to filter out spurious ones, recreating the set when required.
389		421
390	While stopping, setting and starting an I/O watcher in the same iteration	422	While stopping, setting and starting an I/O watcher in the same iteration
391	will result in some caching, there is still a system call per such incident	423	will result in some caching, there is still a system call per such
392	(because the fd could point to a different file description now), so its	424	incident (because the same I<file descriptor> could point to a different
393	best to avoid that. Also, C<dup ()>'ed file descriptors might not work	425	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
394	very well if you register events for both fds.	426	file descriptors might not work very well if you register events for both
395		427	file descriptors.
396	Please note that epoll sometimes generates spurious notifications, so you
397	need to use non-blocking I/O or other means to avoid blocking when no data
398	(or space) is available.
399		428
400	Best performance from this backend is achieved by not unregistering all	429	Best performance from this backend is achieved by not unregistering all
401	watchers for a file descriptor until it has been closed, if possible,	430	watchers for a file descriptor until it has been closed, if possible,
402	i.e. keep at least one watcher active per fd at all times. Stopping and	431	i.e. keep at least one watcher active per fd at all times. Stopping and
403	starting a watcher (without re-setting it) also usually doesn't cause	432	starting a watcher (without re-setting it) also usually doesn't cause
404	extra overhead.	433	extra overhead. A fork can both result in spurious notifications as well
		434	as in libev having to destroy and recreate the epoll object, which can
		435	take considerable time and thus should be avoided.
		436
		437	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		438	faster than epoll for maybe up to a hundred file descriptors, depending on
		439	the usage. So sad.
405		440
406	While nominally embeddable in other event loops, this feature is broken in	441	While nominally embeddable in other event loops, this feature is broken in
407	all kernel versions tested so far.	442	all kernel versions tested so far.
408		443
409	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	444	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
410	C<EVBACKEND_POLL>.	445	C<EVBACKEND_POLL>.
411		446
412	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)	447	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
413		448
414	Kqueue deserves special mention, as at the time of this writing, it was	449	Kqueue deserves special mention, as at the time of this writing, it
415	broken on all BSDs except NetBSD (usually it doesn't work reliably with	450	was broken on all BSDs except NetBSD (usually it doesn't work reliably
416	anything but sockets and pipes, except on Darwin, where of course it's	451	with anything but sockets and pipes, except on Darwin, where of course
417	completely useless). For this reason it's not being "auto-detected" unless	452	it's completely useless). Unlike epoll, however, whose brokenness
418	you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or	453	is by design, these kqueue bugs can (and eventually will) be fixed
419	libev was compiled on a known-to-be-good (-enough) system like NetBSD.	454	without API changes to existing programs. For this reason it's not being
		455	"auto-detected" unless you explicitly specify it in the flags (i.e. using
		456	C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough)
		457	system like NetBSD.
420		458
421	You still can embed kqueue into a normal poll or select backend and use it	459	You still can embed kqueue into a normal poll or select backend and use it
422	only for sockets (after having made sure that sockets work with kqueue on	460	only for sockets (after having made sure that sockets work with kqueue on
423	the target platform). See C<ev_embed> watchers for more info.	461	the target platform). See C<ev_embed> watchers for more info.
424		462
425	It scales in the same way as the epoll backend, but the interface to the	463	It scales in the same way as the epoll backend, but the interface to the
426	kernel is more efficient (which says nothing about its actual speed, of	464	kernel is more efficient (which says nothing about its actual speed, of
427	course). While stopping, setting and starting an I/O watcher does never	465	course). While stopping, setting and starting an I/O watcher does never
428	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to	466	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
429	two event changes per incident. Support for C<fork ()> is very bad and it	467	two event changes per incident. Support for C<fork ()> is very bad (but
430	drops fds silently in similarly hard-to-detect cases.	468	sane, unlike epoll) and it drops fds silently in similarly hard-to-detect
		469	cases
431		470
432	This backend usually performs well under most conditions.	471	This backend usually performs well under most conditions.
433		472
434	While nominally embeddable in other event loops, this doesn't work	473	While nominally embeddable in other event loops, this doesn't work
435	everywhere, so you might need to test for this. And since it is broken	474	everywhere, so you might need to test for this. And since it is broken
436	almost everywhere, you should only use it when you have a lot of sockets	475	almost everywhere, you should only use it when you have a lot of sockets
437	(for which it usually works), by embedding it into another event loop	476	(for which it usually works), by embedding it into another event loop
438	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>) and, did I mention it,	477	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
439	using it only for sockets.	478	also broken on OS X)) and, did I mention it, using it only for sockets.
440		479
441	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with	480	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
442	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with	481	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
443	C<NOTE_EOF>.	482	C<NOTE_EOF>.
444		483
…		…
464	might perform better.	503	might perform better.
465		504
466	On the positive side, with the exception of the spurious readiness	505	On the positive side, with the exception of the spurious readiness
467	notifications, this backend actually performed fully to specification	506	notifications, this backend actually performed fully to specification
468	in all tests and is fully embeddable, which is a rare feat among the	507	in all tests and is fully embeddable, which is a rare feat among the
469	OS-specific backends.	508	OS-specific backends (I vastly prefer correctness over speed hacks).
470		509
471	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as	510	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
472	C<EVBACKEND_POLL>.	511	C<EVBACKEND_POLL>.
473		512
474	=item C<EVBACKEND_ALL>	513	=item C<EVBACKEND_ALL>
…		…
527	responsibility to either stop all watchers cleanly yourself I<before>	566	responsibility to either stop all watchers cleanly yourself I<before>
528	calling this function, or cope with the fact afterwards (which is usually	567	calling this function, or cope with the fact afterwards (which is usually
529	the easiest thing, you can just ignore the watchers and/or C<free ()> them	568	the easiest thing, you can just ignore the watchers and/or C<free ()> them
530	for example).	569	for example).
531		570
532	Note that certain global state, such as signal state, will not be freed by	571	Note that certain global state, such as signal state (and installed signal
533	this function, and related watchers (such as signal and child watchers)	572	handlers), will not be freed by this function, and related watchers (such
534	would need to be stopped manually.	573	as signal and child watchers) would need to be stopped manually.
535		574
536	In general it is not advisable to call this function except in the	575	In general it is not advisable to call this function except in the
537	rare occasion where you really need to free e.g. the signal handling	576	rare occasion where you really need to free e.g. the signal handling
538	pipe fds. If you need dynamically allocated loops it is better to use	577	pipe fds. If you need dynamically allocated loops it is better to use
539	C<ev_loop_new> and C<ev_loop_destroy>).	578	C<ev_loop_new> and C<ev_loop_destroy>).
…		…
605		644
606	This function is rarely useful, but when some event callback runs for a	645	This function is rarely useful, but when some event callback runs for a
607	very long time without entering the event loop, updating libev's idea of	646	very long time without entering the event loop, updating libev's idea of
608	the current time is a good idea.	647	the current time is a good idea.
609		648
610	See also "The special problem of time updates" in the C<ev_timer> section.	649	See also L<The special problem of time updates> in the C<ev_timer> section.
		650
		651	=item ev_suspend (loop)
		652
		653	=item ev_resume (loop)
		654
		655	These two functions suspend and resume a loop, for use when the loop is
		656	not used for a while and timeouts should not be processed.
		657
		658	A typical use case would be an interactive program such as a game: When
		659	the user presses C<^Z> to suspend the game and resumes it an hour later it
		660	would be best to handle timeouts as if no time had actually passed while
		661	the program was suspended. This can be achieved by calling C<ev_suspend>
		662	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		663	C<ev_resume> directly afterwards to resume timer processing.
		664
		665	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		666	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		667	will be rescheduled (that is, they will lose any events that would have
		668	occured while suspended).
		669
		670	After calling C<ev_suspend> you B<must not> call I<any> function on the
		671	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		672	without a previous call to C<ev_suspend>.
		673
		674	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		675	event loop time (see C<ev_now_update>).
611		676
612	=item ev_loop (loop, int flags)	677	=item ev_loop (loop, int flags)
613		678
614	Finally, this is it, the event handler. This function usually is called	679	Finally, this is it, the event handler. This function usually is called
615	after you initialised all your watchers and you want to start handling	680	after you initialised all your watchers and you want to start handling
…		…
631	the loop.	696	the loop.
632		697
633	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	698	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if
634	necessary) and will handle those and any already outstanding ones. It	699	necessary) and will handle those and any already outstanding ones. It
635	will block your process until at least one new event arrives (which could	700	will block your process until at least one new event arrives (which could
636	be an event internal to libev itself, so there is no guarentee that a	701	be an event internal to libev itself, so there is no guarantee that a
637	user-registered callback will be called), and will return after one	702	user-registered callback will be called), and will return after one
638	iteration of the loop.	703	iteration of the loop.
639		704
640	This is useful if you are waiting for some external event in conjunction	705	This is useful if you are waiting for some external event in conjunction
641	with something not expressible using other libev watchers (i.e. "roll your	706	with something not expressible using other libev watchers (i.e. "roll your
…		…
699		764
700	If you have a watcher you never unregister that should not keep C<ev_loop>	765	If you have a watcher you never unregister that should not keep C<ev_loop>
701	from returning, call ev_unref() after starting, and ev_ref() before	766	from returning, call ev_unref() after starting, and ev_ref() before
702	stopping it.	767	stopping it.
703		768
704	As an example, libev itself uses this for its internal signal pipe: It is	769	As an example, libev itself uses this for its internal signal pipe: It
705	not visible to the libev user and should not keep C<ev_loop> from exiting	770	is not visible to the libev user and should not keep C<ev_loop> from
706	if no event watchers registered by it are active. It is also an excellent	771	exiting if no event watchers registered by it are active. It is also an
707	way to do this for generic recurring timers or from within third-party	772	excellent way to do this for generic recurring timers or from within
708	libraries. Just remember to I<unref after start> and I<ref before stop>	773	third-party libraries. Just remember to I<unref after start> and I<ref
709	(but only if the watcher wasn't active before, or was active before,	774	before stop> (but only if the watcher wasn't active before, or was active
710	respectively).	775	before, respectively. Note also that libev might stop watchers itself
		776	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		777	in the callback).
711		778
712	Example: Create a signal watcher, but keep it from keeping C<ev_loop>	779	Example: Create a signal watcher, but keep it from keeping C<ev_loop>
713	running when nothing else is active.	780	running when nothing else is active.
714		781
715	ev_signal exitsig;	782	ev_signal exitsig;
…		…
744		811
745	By setting a higher I<io collect interval> you allow libev to spend more	812	By setting a higher I<io collect interval> you allow libev to spend more
746	time collecting I/O events, so you can handle more events per iteration,	813	time collecting I/O events, so you can handle more events per iteration,
747	at the cost of increasing latency. Timeouts (both C<ev_periodic> and	814	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
748	C<ev_timer>) will be not affected. Setting this to a non-null value will	815	C<ev_timer>) will be not affected. Setting this to a non-null value will
749	introduce an additional C<ev_sleep ()> call into most loop iterations.	816	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		817	sleep time ensures that libev will not poll for I/O events more often then
		818	once per this interval, on average.
750		819
751	Likewise, by setting a higher I<timeout collect interval> you allow libev	820	Likewise, by setting a higher I<timeout collect interval> you allow libev
752	to spend more time collecting timeouts, at the expense of increased	821	to spend more time collecting timeouts, at the expense of increased
753	latency/jitter/inexactness (the watcher callback will be called	822	latency/jitter/inexactness (the watcher callback will be called
754	later). C<ev_io> watchers will not be affected. Setting this to a non-null	823	later). C<ev_io> watchers will not be affected. Setting this to a non-null
…		…
756		825
757	Many (busy) programs can usually benefit by setting the I/O collect	826	Many (busy) programs can usually benefit by setting the I/O collect
758	interval to a value near C<0.1> or so, which is often enough for	827	interval to a value near C<0.1> or so, which is often enough for
759	interactive servers (of course not for games), likewise for timeouts. It	828	interactive servers (of course not for games), likewise for timeouts. It
760	usually doesn't make much sense to set it to a lower value than C<0.01>,	829	usually doesn't make much sense to set it to a lower value than C<0.01>,
761	as this approaches the timing granularity of most systems.	830	as this approaches the timing granularity of most systems. Note that if
		831	you do transactions with the outside world and you can't increase the
		832	parallelity, then this setting will limit your transaction rate (if you
		833	need to poll once per transaction and the I/O collect interval is 0.01,
		834	then you can't do more than 100 transations per second).
762		835
763	Setting the I<timeout collect interval> can improve the opportunity for	836	Setting the I<timeout collect interval> can improve the opportunity for
764	saving power, as the program will "bundle" timer callback invocations that	837	saving power, as the program will "bundle" timer callback invocations that
765	are "near" in time together, by delaying some, thus reducing the number of	838	are "near" in time together, by delaying some, thus reducing the number of
766	times the process sleeps and wakes up again. Another useful technique to	839	times the process sleeps and wakes up again. Another useful technique to
767	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure	840	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
768	they fire on, say, one-second boundaries only.	841	they fire on, say, one-second boundaries only.
769		842
		843	Example: we only need 0.1s timeout granularity, and we wish not to poll
		844	more often than 100 times per second:
		845
		846	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		847	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		848
770	=item ev_loop_verify (loop)	849	=item ev_loop_verify (loop)
771		850
772	This function only does something when C<EV_VERIFY> support has been	851	This function only does something when C<EV_VERIFY> support has been
773	compiled in. which is the default for non-minimal builds. It tries to go	852	compiled in, which is the default for non-minimal builds. It tries to go
774	through all internal structures and checks them for validity. If anything	853	through all internal structures and checks them for validity. If anything
775	is found to be inconsistent, it will print an error message to standard	854	is found to be inconsistent, it will print an error message to standard
776	error and call C<abort ()>.	855	error and call C<abort ()>.
777		856
778	This can be used to catch bugs inside libev itself: under normal	857	This can be used to catch bugs inside libev itself: under normal
…		…
781		860
782	=back	861	=back
783		862
784		863
785	=head1 ANATOMY OF A WATCHER	864	=head1 ANATOMY OF A WATCHER
		865
		866	In the following description, uppercase C<TYPE> in names stands for the
		867	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		868	watchers and C<ev_io_start> for I/O watchers.
786		869
787	A watcher is a structure that you create and register to record your	870	A watcher is a structure that you create and register to record your
788	interest in some event. For instance, if you want to wait for STDIN to	871	interest in some event. For instance, if you want to wait for STDIN to
789	become readable, you would create an C<ev_io> watcher for that:	872	become readable, you would create an C<ev_io> watcher for that:
790		873
…		…
793	ev_io_stop (w);	876	ev_io_stop (w);
794	ev_unloop (loop, EVUNLOOP_ALL);	877	ev_unloop (loop, EVUNLOOP_ALL);
795	}	878	}
796		879
797	struct ev_loop *loop = ev_default_loop (0);	880	struct ev_loop *loop = ev_default_loop (0);
		881
798	ev_io stdin_watcher;	882	ev_io stdin_watcher;
		883
799	ev_init (&stdin_watcher, my_cb);	884	ev_init (&stdin_watcher, my_cb);
800	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	885	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
801	ev_io_start (loop, &stdin_watcher);	886	ev_io_start (loop, &stdin_watcher);
		887
802	ev_loop (loop, 0);	888	ev_loop (loop, 0);
803		889
804	As you can see, you are responsible for allocating the memory for your	890	As you can see, you are responsible for allocating the memory for your
805	watcher structures (and it is usually a bad idea to do this on the stack,	891	watcher structures (and it is I<usually> a bad idea to do this on the
806	although this can sometimes be quite valid).	892	stack).
		893
		894	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		895	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
807		896
808	Each watcher structure must be initialised by a call to C<ev_init	897	Each watcher structure must be initialised by a call to C<ev_init
809	(watcher *, callback)>, which expects a callback to be provided. This	898	(watcher *, callback)>, which expects a callback to be provided. This
810	callback gets invoked each time the event occurs (or, in the case of I/O	899	callback gets invoked each time the event occurs (or, in the case of I/O
811	watchers, each time the event loop detects that the file descriptor given	900	watchers, each time the event loop detects that the file descriptor given
812	is readable and/or writable).	901	is readable and/or writable).
813		902
814	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	903	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
815	with arguments specific to this watcher type. There is also a macro	904	macro to configure it, with arguments specific to the watcher type. There
816	to combine initialisation and setting in one call: C<< ev_<type>_init	905	is also a macro to combine initialisation and setting in one call: C<<
817	(watcher *, callback, ...) >>.	906	ev_TYPE_init (watcher *, callback, ...) >>.
818		907
819	To make the watcher actually watch out for events, you have to start it	908	To make the watcher actually watch out for events, you have to start it
820	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	909	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
821	*) >>), and you can stop watching for events at any time by calling the	910	*) >>), and you can stop watching for events at any time by calling the
822	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	911	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
823		912
824	As long as your watcher is active (has been started but not stopped) you	913	As long as your watcher is active (has been started but not stopped) you
825	must not touch the values stored in it. Most specifically you must never	914	must not touch the values stored in it. Most specifically you must never
826	reinitialise it or call its C<set> macro.	915	reinitialise it or call its C<ev_TYPE_set> macro.
827		916
828	Each and every callback receives the event loop pointer as first, the	917	Each and every callback receives the event loop pointer as first, the
829	registered watcher structure as second, and a bitset of received events as	918	registered watcher structure as second, and a bitset of received events as
830	third argument.	919	third argument.
831		920
…		…
889		978
890	=item C<EV_ASYNC>	979	=item C<EV_ASYNC>
891		980
892	The given async watcher has been asynchronously notified (see C<ev_async>).	981	The given async watcher has been asynchronously notified (see C<ev_async>).
893		982
		983	=item C<EV_CUSTOM>
		984
		985	Not ever sent (or otherwise used) by libev itself, but can be freely used
		986	by libev users to signal watchers (e.g. via C<ev_feed_event>).
		987
894	=item C<EV_ERROR>	988	=item C<EV_ERROR>
895		989
896	An unspecified error has occurred, the watcher has been stopped. This might	990	An unspecified error has occurred, the watcher has been stopped. This might
897	happen because the watcher could not be properly started because libev	991	happen because the watcher could not be properly started because libev
898	ran out of memory, a file descriptor was found to be closed or any other	992	ran out of memory, a file descriptor was found to be closed or any other
…		…
912		1006
913	=back	1007	=back
914		1008
915	=head2 GENERIC WATCHER FUNCTIONS	1009	=head2 GENERIC WATCHER FUNCTIONS
916		1010
917	In the following description, C<TYPE> stands for the watcher type,
918	e.g. C<timer> for C<ev_timer> watchers and C<io> for C<ev_io> watchers.
919
920	=over 4	1011	=over 4
921		1012
922	=item C<ev_init> (ev_TYPE *watcher, callback)	1013	=item C<ev_init> (ev_TYPE *watcher, callback)
923		1014
924	This macro initialises the generic portion of a watcher. The contents	1015	This macro initialises the generic portion of a watcher. The contents
…		…
1016	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>	1107	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
1017	(default: C<-2>). Pending watchers with higher priority will be invoked	1108	(default: C<-2>). Pending watchers with higher priority will be invoked
1018	before watchers with lower priority, but priority will not keep watchers	1109	before watchers with lower priority, but priority will not keep watchers
1019	from being executed (except for C<ev_idle> watchers).	1110	from being executed (except for C<ev_idle> watchers).
1020		1111
1021	This means that priorities are I<only> used for ordering callback
1022	invocation after new events have been received. This is useful, for
1023	example, to reduce latency after idling, or more often, to bind two
1024	watchers on the same event and make sure one is called first.
1025
1026	If you need to suppress invocation when higher priority events are pending	1112	If you need to suppress invocation when higher priority events are pending
1027	you need to look at C<ev_idle> watchers, which provide this functionality.	1113	you need to look at C<ev_idle> watchers, which provide this functionality.
1028		1114
1029	You I<must not> change the priority of a watcher as long as it is active or	1115	You I<must not> change the priority of a watcher as long as it is active or
1030	pending.	1116	pending.
1031		1117
		1118	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1119	fine, as long as you do not mind that the priority value you query might
		1120	or might not have been clamped to the valid range.
		1121
1032	The default priority used by watchers when no priority has been set is	1122	The default priority used by watchers when no priority has been set is
1033	always C<0>, which is supposed to not be too high and not be too low :).	1123	always C<0>, which is supposed to not be too high and not be too low :).
1034		1124
1035	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is	1125	See L<WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
1036	fine, as long as you do not mind that the priority value you query might	1126	priorities.
1037	or might not have been adjusted to be within valid range.
1038		1127
1039	=item ev_invoke (loop, ev_TYPE *watcher, int revents)	1128	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
1040		1129
1041	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither	1130	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
1042	C<loop> nor C<revents> need to be valid as long as the watcher callback	1131	C<loop> nor C<revents> need to be valid as long as the watcher callback
…		…
1107	#include <stddef.h>	1196	#include <stddef.h>
1108		1197
1109	static void	1198	static void
1110	t1_cb (EV_P_ ev_timer *w, int revents)	1199	t1_cb (EV_P_ ev_timer *w, int revents)
1111	{	1200	{
1112	struct my_biggy big = (struct my_biggy *	1201	struct my_biggy big = (struct my_biggy *)
1113	(((char *)w) - offsetof (struct my_biggy, t1));	1202	(((char *)w) - offsetof (struct my_biggy, t1));
1114	}	1203	}
1115		1204
1116	static void	1205	static void
1117	t2_cb (EV_P_ ev_timer *w, int revents)	1206	t2_cb (EV_P_ ev_timer *w, int revents)
1118	{	1207	{
1119	struct my_biggy big = (struct my_biggy *	1208	struct my_biggy big = (struct my_biggy *)
1120	(((char *)w) - offsetof (struct my_biggy, t2));	1209	(((char *)w) - offsetof (struct my_biggy, t2));
1121	}	1210	}
		1211
		1212	=head2 WATCHER PRIORITY MODELS
		1213
		1214	Many event loops support I<watcher priorities>, which are usually small
		1215	integers that influence the ordering of event callback invocation
		1216	between watchers in some way, all else being equal.
		1217
		1218	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1219	description for the more technical details such as the actual priority
		1220	range.
		1221
		1222	There are two common ways how these these priorities are being interpreted
		1223	by event loops:
		1224
		1225	In the more common lock-out model, higher priorities "lock out" invocation
		1226	of lower priority watchers, which means as long as higher priority
		1227	watchers receive events, lower priority watchers are not being invoked.
		1228
		1229	The less common only-for-ordering model uses priorities solely to order
		1230	callback invocation within a single event loop iteration: Higher priority
		1231	watchers are invoked before lower priority ones, but they all get invoked
		1232	before polling for new events.
		1233
		1234	Libev uses the second (only-for-ordering) model for all its watchers
		1235	except for idle watchers (which use the lock-out model).
		1236
		1237	The rationale behind this is that implementing the lock-out model for
		1238	watchers is not well supported by most kernel interfaces, and most event
		1239	libraries will just poll for the same events again and again as long as
		1240	their callbacks have not been executed, which is very inefficient in the
		1241	common case of one high-priority watcher locking out a mass of lower
		1242	priority ones.
		1243
		1244	Static (ordering) priorities are most useful when you have two or more
		1245	watchers handling the same resource: a typical usage example is having an
		1246	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1247	timeouts. Under load, data might be received while the program handles
		1248	other jobs, but since timers normally get invoked first, the timeout
		1249	handler will be executed before checking for data. In that case, giving
		1250	the timer a lower priority than the I/O watcher ensures that I/O will be
		1251	handled first even under adverse conditions (which is usually, but not
		1252	always, what you want).
		1253
		1254	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1255	will only be executed when no same or higher priority watchers have
		1256	received events, they can be used to implement the "lock-out" model when
		1257	required.
		1258
		1259	For example, to emulate how many other event libraries handle priorities,
		1260	you can associate an C<ev_idle> watcher to each such watcher, and in
		1261	the normal watcher callback, you just start the idle watcher. The real
		1262	processing is done in the idle watcher callback. This causes libev to
		1263	continously poll and process kernel event data for the watcher, but when
		1264	the lock-out case is known to be rare (which in turn is rare :), this is
		1265	workable.
		1266
		1267	Usually, however, the lock-out model implemented that way will perform
		1268	miserably under the type of load it was designed to handle. In that case,
		1269	it might be preferable to stop the real watcher before starting the
		1270	idle watcher, so the kernel will not have to process the event in case
		1271	the actual processing will be delayed for considerable time.
		1272
		1273	Here is an example of an I/O watcher that should run at a strictly lower
		1274	priority than the default, and which should only process data when no
		1275	other events are pending:
		1276
		1277	ev_idle idle; // actual processing watcher
		1278	ev_io io; // actual event watcher
		1279
		1280	static void
		1281	io_cb (EV_P_ ev_io *w, int revents)
		1282	{
		1283	// stop the I/O watcher, we received the event, but
		1284	// are not yet ready to handle it.
		1285	ev_io_stop (EV_A_ w);
		1286
		1287	// start the idle watcher to ahndle the actual event.
		1288	// it will not be executed as long as other watchers
		1289	// with the default priority are receiving events.
		1290	ev_idle_start (EV_A_ &idle);
		1291	}
		1292
		1293	static void
		1294	idle_cb (EV_P_ ev_idle *w, int revents)
		1295	{
		1296	// actual processing
		1297	read (STDIN_FILENO, ...);
		1298
		1299	// have to start the I/O watcher again, as
		1300	// we have handled the event
		1301	ev_io_start (EV_P_ &io);
		1302	}
		1303
		1304	// initialisation
		1305	ev_idle_init (&idle, idle_cb);
		1306	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1307	ev_io_start (EV_DEFAULT_ &io);
		1308
		1309	In the "real" world, it might also be beneficial to start a timer, so that
		1310	low-priority connections can not be locked out forever under load. This
		1311	enables your program to keep a lower latency for important connections
		1312	during short periods of high load, while not completely locking out less
		1313	important ones.
1122		1314
1123		1315
1124	=head1 WATCHER TYPES	1316	=head1 WATCHER TYPES
1125		1317
1126	This section describes each watcher in detail, but will not repeat	1318	This section describes each watcher in detail, but will not repeat
…		…
1152	descriptors to non-blocking mode is also usually a good idea (but not	1344	descriptors to non-blocking mode is also usually a good idea (but not
1153	required if you know what you are doing).	1345	required if you know what you are doing).
1154		1346
1155	If you cannot use non-blocking mode, then force the use of a	1347	If you cannot use non-blocking mode, then force the use of a
1156	known-to-be-good backend (at the time of this writing, this includes only	1348	known-to-be-good backend (at the time of this writing, this includes only
1157	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>).	1349	C<EVBACKEND_SELECT> and C<EVBACKEND_POLL>). The same applies to file
		1350	descriptors for which non-blocking operation makes no sense (such as
		1351	files) - libev doesn't guarentee any specific behaviour in that case.
1158		1352
1159	Another thing you have to watch out for is that it is quite easy to	1353	Another thing you have to watch out for is that it is quite easy to
1160	receive "spurious" readiness notifications, that is your callback might	1354	receive "spurious" readiness notifications, that is your callback might
1161	be called with C<EV_READ> but a subsequent C<read>(2) will actually block	1355	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
1162	because there is no data. Not only are some backends known to create a	1356	because there is no data. Not only are some backends known to create a
…		…
1283	year, it will still time out after (roughly) one hour. "Roughly" because	1477	year, it will still time out after (roughly) one hour. "Roughly" because
1284	detecting time jumps is hard, and some inaccuracies are unavoidable (the	1478	detecting time jumps is hard, and some inaccuracies are unavoidable (the
1285	monotonic clock option helps a lot here).	1479	monotonic clock option helps a lot here).
1286		1480
1287	The callback is guaranteed to be invoked only I<after> its timeout has	1481	The callback is guaranteed to be invoked only I<after> its timeout has
1288	passed, but if multiple timers become ready during the same loop iteration	1482	passed (not I<at>, so on systems with very low-resolution clocks this
1289	then order of execution is undefined.	1483	might introduce a small delay). If multiple timers become ready during the
		1484	same loop iteration then the ones with earlier time-out values are invoked
		1485	before ones with later time-out values (but this is no longer true when a
		1486	callback calls C<ev_loop> recursively).
1290		1487
1291	=head3 Be smart about timeouts	1488	=head3 Be smart about timeouts
1292		1489
1293	Many real-world problems invole some kind of time-out, usually for error	1490	Many real-world problems involve some kind of timeout, usually for error
1294	recovery. A typical example is an HTTP request - if the other side hangs,	1491	recovery. A typical example is an HTTP request - if the other side hangs,
1295	you want to raise some error after a while.	1492	you want to raise some error after a while.
1296		1493
1297	Here are some ways on how to handle this problem, from simple and	1494	What follows are some ways to handle this problem, from obvious and
1298	inefficient to very efficient.	1495	inefficient to smart and efficient.
1299		1496
1300	In the following examples a 60 second activity timeout is assumed - a	1497	In the following, a 60 second activity timeout is assumed - a timeout that
1301	timeout that gets reset to 60 seconds each time some data ("a lifesign")	1498	gets reset to 60 seconds each time there is activity (e.g. each time some
1302	was received.	1499	data or other life sign was received).
1303		1500
1304	=over 4	1501	=over 4
1305		1502
1306	=item 1. Use a timer and stop, reinitialise, start it on activity.	1503	=item 1. Use a timer and stop, reinitialise and start it on activity.
1307		1504
1308	This is the most obvious, but not the most simple way: In the beginning,	1505	This is the most obvious, but not the most simple way: In the beginning,
1309	start the watcher:	1506	start the watcher:
1310		1507
1311	ev_timer_init (timer, callback, 60., 0.);	1508	ev_timer_init (timer, callback, 60., 0.);
1312	ev_timer_start (loop, timer);	1509	ev_timer_start (loop, timer);
1313		1510
1314	Then, each time there is some activity, C<ev_timer_stop> the timer,	1511	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
1315	initialise it again, and start it:	1512	and start it again:
1316		1513
1317	ev_timer_stop (loop, timer);	1514	ev_timer_stop (loop, timer);
1318	ev_timer_set (timer, 60., 0.);	1515	ev_timer_set (timer, 60., 0.);
1319	ev_timer_start (loop, timer);	1516	ev_timer_start (loop, timer);
1320		1517
1321	This is relatively simple to implement, but means that each time there	1518	This is relatively simple to implement, but means that each time there is
1322	is some activity, libev will first have to remove the timer from it's	1519	some activity, libev will first have to remove the timer from its internal
1323	internal data strcuture and then add it again.	1520	data structure and then add it again. Libev tries to be fast, but it's
		1521	still not a constant-time operation.
1324		1522
1325	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.	1523	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
1326		1524
1327	This is the easiest way, and involves using C<ev_timer_again> instead of	1525	This is the easiest way, and involves using C<ev_timer_again> instead of
1328	C<ev_timer_start>.	1526	C<ev_timer_start>.
1329		1527
1330	For this, configure an C<ev_timer> with a C<repeat> value of C<60> and	1528	To implement this, configure an C<ev_timer> with a C<repeat> value
1331	then call C<ev_timer_again> at start and each time you successfully read	1529	of C<60> and then call C<ev_timer_again> at start and each time you
1332	or write some data. If you go into an idle state where you do not expect	1530	successfully read or write some data. If you go into an idle state where
1333	data to travel on the socket, you can C<ev_timer_stop> the timer, and	1531	you do not expect data to travel on the socket, you can C<ev_timer_stop>
1334	C<ev_timer_again> will automatically restart it if need be.	1532	the timer, and C<ev_timer_again> will automatically restart it if need be.
1335		1533
1336	That means you can ignore the C<after> value and C<ev_timer_start>	1534	That means you can ignore both the C<ev_timer_start> function and the
1337	altogether and only ever use the C<repeat> value and C<ev_timer_again>.	1535	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1536	member and C<ev_timer_again>.
1338		1537
1339	At start:	1538	At start:
1340		1539
1341	ev_timer_init (timer, callback, 0., 60.);	1540	ev_init (timer, callback);
		1541	timer->repeat = 60.;
1342	ev_timer_again (loop, timer);	1542	ev_timer_again (loop, timer);
1343		1543
1344	Each time you receive some data:	1544	Each time there is some activity:
1345		1545
1346	ev_timer_again (loop, timer);	1546	ev_timer_again (loop, timer);
1347		1547
1348	It is even possible to change the time-out on the fly:	1548	It is even possible to change the time-out on the fly, regardless of
		1549	whether the watcher is active or not:
1349		1550
1350	timer->repeat = 30.;	1551	timer->repeat = 30.;
1351	ev_timer_again (loop, timer);	1552	ev_timer_again (loop, timer);
1352		1553
1353	This is slightly more efficient then stopping/starting the timer each time	1554	This is slightly more efficient then stopping/starting the timer each time
1354	you want to modify its timeout value, as libev does not have to completely	1555	you want to modify its timeout value, as libev does not have to completely
1355	remove and re-insert the timer from/into it's internal data structure.	1556	remove and re-insert the timer from/into its internal data structure.
		1557
		1558	It is, however, even simpler than the "obvious" way to do it.
1356		1559
1357	=item 3. Let the timer time out, but then re-arm it as required.	1560	=item 3. Let the timer time out, but then re-arm it as required.
1358		1561
1359	This method is more tricky, but usually most efficient: Most timeouts are	1562	This method is more tricky, but usually most efficient: Most timeouts are
1360	relatively long compared to the loop iteration time - in our example,	1563	relatively long compared to the intervals between other activity - in
1361	within 60 seconds, there are usually many I/O events with associated	1564	our example, within 60 seconds, there are usually many I/O events with
1362	activity resets.	1565	associated activity resets.
1363		1566
1364	In this case, it would be more efficient to leave the C<ev_timer> alone,	1567	In this case, it would be more efficient to leave the C<ev_timer> alone,
1365	but remember the time of last activity, and check for a real timeout only	1568	but remember the time of last activity, and check for a real timeout only
1366	within the callback:	1569	within the callback:
1367		1570
1368	ev_tstamp last_activity; // time of last activity	1571	ev_tstamp last_activity; // time of last activity
1369		1572
1370	static void	1573	static void
1371	callback (EV_P_ ev_timer *w, int revents)	1574	callback (EV_P_ ev_timer *w, int revents)
1372	{	1575	{
1373	ev_tstamp now = ev_now (EV_A);	1576	ev_tstamp now = ev_now (EV_A);
1374	ev_tstamp timeout = last_activity + 60.;	1577	ev_tstamp timeout = last_activity + 60.;
1375		1578
1376	// if last_activity is older than now - timeout, we did time out	1579	// if last_activity + 60. is older than now, we did time out
1377	if (timeout < now)	1580	if (timeout < now)
1378	{	1581	{
1379	// timeout occured, take action	1582	// timeout occured, take action
1380	}	1583	}
1381	else	1584	else
1382	{	1585	{
1383	// callback was invoked, but there was some activity, re-arm	1586	// callback was invoked, but there was some activity, re-arm
1384	// to fire in last_activity + 60.	1587	// the watcher to fire in last_activity + 60, which is
		1588	// guaranteed to be in the future, so "again" is positive:
1385	w->again = timeout - now;	1589	w->repeat = timeout - now;
1386	ev_timer_again (EV_A_ w);	1590	ev_timer_again (EV_A_ w);
1387	}	1591	}
1388	}	1592	}
1389		1593
1390	To summarise the callback: first calculate the real time-out (defined as	1594	To summarise the callback: first calculate the real timeout (defined
1391	"60 seconds after the last activity"), then check if that time has been	1595	as "60 seconds after the last activity"), then check if that time has
1392	reached, which means there was a real timeout. Otherwise the callback was	1596	been reached, which means something I<did>, in fact, time out. Otherwise
1393	invoked too early (timeout is in the future), so re-schedule the timer to	1597	the callback was invoked too early (C<timeout> is in the future), so
1394	fire at that future time.	1598	re-schedule the timer to fire at that future time, to see if maybe we have
		1599	a timeout then.
1395		1600
1396	Note how C<ev_timer_again> is used, taking advantage of the	1601	Note how C<ev_timer_again> is used, taking advantage of the
1397	C<ev_timer_again> optimisation when the timer is already running.	1602	C<ev_timer_again> optimisation when the timer is already running.
1398		1603
1399	This scheme causes more callback invocations (about one every 60 seconds),	1604	This scheme causes more callback invocations (about one every 60 seconds
1400	but virtually no calls to libev to change the timeout.	1605	minus half the average time between activity), but virtually no calls to
		1606	libev to change the timeout.
1401		1607
1402	To start the timer, simply intiialise the watcher and C<last_activity>,	1608	To start the timer, simply initialise the watcher and set C<last_activity>
1403	then call the callback:	1609	to the current time (meaning we just have some activity :), then call the
		1610	callback, which will "do the right thing" and start the timer:
1404		1611
1405	ev_timer_init (timer, callback);	1612	ev_init (timer, callback);
1406	last_activity = ev_now (loop);	1613	last_activity = ev_now (loop);
1407	callback (loop, timer, EV_TIMEOUT);	1614	callback (loop, timer, EV_TIMEOUT);
1408		1615
1409	And when there is some activity, simply remember the time in	1616	And when there is some activity, simply store the current time in
1410	C<last_activity>:	1617	C<last_activity>, no libev calls at all:
1411		1618
1412	last_actiivty = ev_now (loop);	1619	last_actiivty = ev_now (loop);
1413		1620
1414	This technique is slightly more complex, but in most cases where the	1621	This technique is slightly more complex, but in most cases where the
1415	time-out is unlikely to be triggered, much more efficient.	1622	time-out is unlikely to be triggered, much more efficient.
1416		1623
		1624	Changing the timeout is trivial as well (if it isn't hard-coded in the
		1625	callback :) - just change the timeout and invoke the callback, which will
		1626	fix things for you.
		1627
		1628	=item 4. Wee, just use a double-linked list for your timeouts.
		1629
		1630	If there is not one request, but many thousands (millions...), all
		1631	employing some kind of timeout with the same timeout value, then one can
		1632	do even better:
		1633
		1634	When starting the timeout, calculate the timeout value and put the timeout
		1635	at the I<end> of the list.
		1636
		1637	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		1638	the list is expected to fire (for example, using the technique #3).
		1639
		1640	When there is some activity, remove the timer from the list, recalculate
		1641	the timeout, append it to the end of the list again, and make sure to
		1642	update the C<ev_timer> if it was taken from the beginning of the list.
		1643
		1644	This way, one can manage an unlimited number of timeouts in O(1) time for
		1645	starting, stopping and updating the timers, at the expense of a major
		1646	complication, and having to use a constant timeout. The constant timeout
		1647	ensures that the list stays sorted.
		1648
1417	=back	1649	=back
		1650
		1651	So which method the best?
		1652
		1653	Method #2 is a simple no-brain-required solution that is adequate in most
		1654	situations. Method #3 requires a bit more thinking, but handles many cases
		1655	better, and isn't very complicated either. In most case, choosing either
		1656	one is fine, with #3 being better in typical situations.
		1657
		1658	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		1659	rather complicated, but extremely efficient, something that really pays
		1660	off after the first million or so of active timers, i.e. it's usually
		1661	overkill :)
1418		1662
1419	=head3 The special problem of time updates	1663	=head3 The special problem of time updates
1420		1664
1421	Establishing the current time is a costly operation (it usually takes at	1665	Establishing the current time is a costly operation (it usually takes at
1422	least two system calls): EV therefore updates its idea of the current	1666	least two system calls): EV therefore updates its idea of the current
…		…
1466	If the timer is started but non-repeating, stop it (as if it timed out).	1710	If the timer is started but non-repeating, stop it (as if it timed out).
1467		1711
1468	If the timer is repeating, either start it if necessary (with the	1712	If the timer is repeating, either start it if necessary (with the
1469	C<repeat> value), or reset the running timer to the C<repeat> value.	1713	C<repeat> value), or reset the running timer to the C<repeat> value.
1470		1714
1471	This sounds a bit complicated, see "Be smart about timeouts", above, for a	1715	This sounds a bit complicated, see L<Be smart about timeouts>, above, for a
1472	usage example.	1716	usage example.
1473		1717
1474	=item ev_tstamp repeat [read-write]	1718	=item ev_tstamp repeat [read-write]
1475		1719
1476	The current C<repeat> value. Will be used each time the watcher times out	1720	The current C<repeat> value. Will be used each time the watcher times out
…		…
1515	=head2 C<ev_periodic> - to cron or not to cron?	1759	=head2 C<ev_periodic> - to cron or not to cron?
1516		1760
1517	Periodic watchers are also timers of a kind, but they are very versatile	1761	Periodic watchers are also timers of a kind, but they are very versatile
1518	(and unfortunately a bit complex).	1762	(and unfortunately a bit complex).
1519		1763
1520	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	1764	Unlike C<ev_timer>, periodic watchers are not based on real time (or
1521	but on wall clock time (absolute time). You can tell a periodic watcher	1765	relative time, the physical time that passes) but on wall clock time
1522	to trigger after some specific point in time. For example, if you tell a	1766	(absolute time, the thing you can read on your calender or clock). The
1523	periodic watcher to trigger in 10 seconds (by specifying e.g. C<ev_now ()	1767	difference is that wall clock time can run faster or slower than real
1524	+ 10.>, that is, an absolute time not a delay) and then reset your system	1768	time, and time jumps are not uncommon (e.g. when you adjust your
1525	clock to January of the previous year, then it will take more than year	1769	wrist-watch).
1526	to trigger the event (unlike an C<ev_timer>, which would still trigger
1527	roughly 10 seconds later as it uses a relative timeout).
1528		1770
		1771	You can tell a periodic watcher to trigger after some specific point
		1772	in time: for example, if you tell a periodic watcher to trigger "in 10
		1773	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		1774	not a delay) and then reset your system clock to January of the previous
		1775	year, then it will take a year or more to trigger the event (unlike an
		1776	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		1777	it, as it uses a relative timeout).
		1778
1529	C<ev_periodic>s can also be used to implement vastly more complex timers,	1779	C<ev_periodic> watchers can also be used to implement vastly more complex
1530	such as triggering an event on each "midnight, local time", or other	1780	timers, such as triggering an event on each "midnight, local time", or
1531	complicated rules.	1781	other complicated rules. This cannot be done with C<ev_timer> watchers, as
		1782	those cannot react to time jumps.
1532		1783
1533	As with timers, the callback is guaranteed to be invoked only when the	1784	As with timers, the callback is guaranteed to be invoked only when the
1534	time (C<at>) has passed, but if multiple periodic timers become ready	1785	point in time where it is supposed to trigger has passed. If multiple
1535	during the same loop iteration, then order of execution is undefined.	1786	timers become ready during the same loop iteration then the ones with
		1787	earlier time-out values are invoked before ones with later time-out values
		1788	(but this is no longer true when a callback calls C<ev_loop> recursively).
1536		1789
1537	=head3 Watcher-Specific Functions and Data Members	1790	=head3 Watcher-Specific Functions and Data Members
1538		1791
1539	=over 4	1792	=over 4
1540		1793
1541	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	1794	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1542		1795
1543	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	1796	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
1544		1797
1545	Lots of arguments, lets sort it out... There are basically three modes of	1798	Lots of arguments, let's sort it out... There are basically three modes of
1546	operation, and we will explain them from simplest to most complex:	1799	operation, and we will explain them from simplest to most complex:
1547		1800
1548	=over 4	1801	=over 4
1549		1802
1550	=item * absolute timer (at = time, interval = reschedule_cb = 0)	1803	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
1551		1804
1552	In this configuration the watcher triggers an event after the wall clock	1805	In this configuration the watcher triggers an event after the wall clock
1553	time C<at> has passed. It will not repeat and will not adjust when a time	1806	time C<offset> has passed. It will not repeat and will not adjust when a
1554	jump occurs, that is, if it is to be run at January 1st 2011 then it will	1807	time jump occurs, that is, if it is to be run at January 1st 2011 then it
1555	only run when the system clock reaches or surpasses this time.	1808	will be stopped and invoked when the system clock reaches or surpasses
		1809	this point in time.
1556		1810
1557	=item * repeating interval timer (at = offset, interval > 0, reschedule_cb = 0)	1811	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
1558		1812
1559	In this mode the watcher will always be scheduled to time out at the next	1813	In this mode the watcher will always be scheduled to time out at the next
1560	C<at + N * interval> time (for some integer N, which can also be negative)	1814	C<offset + N * interval> time (for some integer N, which can also be
1561	and then repeat, regardless of any time jumps.	1815	negative) and then repeat, regardless of any time jumps. The C<offset>
		1816	argument is merely an offset into the C<interval> periods.
1562		1817
1563	This can be used to create timers that do not drift with respect to the	1818	This can be used to create timers that do not drift with respect to the
1564	system clock, for example, here is a C<ev_periodic> that triggers each	1819	system clock, for example, here is an C<ev_periodic> that triggers each
1565	hour, on the hour:	1820	hour, on the hour (with respect to UTC):
1566		1821
1567	ev_periodic_set (&periodic, 0., 3600., 0);	1822	ev_periodic_set (&periodic, 0., 3600., 0);
1568		1823
1569	This doesn't mean there will always be 3600 seconds in between triggers,	1824	This doesn't mean there will always be 3600 seconds in between triggers,
1570	but only that the callback will be called when the system time shows a	1825	but only that the callback will be called when the system time shows a
1571	full hour (UTC), or more correctly, when the system time is evenly divisible	1826	full hour (UTC), or more correctly, when the system time is evenly divisible
1572	by 3600.	1827	by 3600.
1573		1828
1574	Another way to think about it (for the mathematically inclined) is that	1829	Another way to think about it (for the mathematically inclined) is that
1575	C<ev_periodic> will try to run the callback in this mode at the next possible	1830	C<ev_periodic> will try to run the callback in this mode at the next possible
1576	time where C<time = at (mod interval)>, regardless of any time jumps.	1831	time where C<time = offset (mod interval)>, regardless of any time jumps.
1577		1832
1578	For numerical stability it is preferable that the C<at> value is near	1833	For numerical stability it is preferable that the C<offset> value is near
1579	C<ev_now ()> (the current time), but there is no range requirement for	1834	C<ev_now ()> (the current time), but there is no range requirement for
1580	this value, and in fact is often specified as zero.	1835	this value, and in fact is often specified as zero.
1581		1836
1582	Note also that there is an upper limit to how often a timer can fire (CPU	1837	Note also that there is an upper limit to how often a timer can fire (CPU
1583	speed for example), so if C<interval> is very small then timing stability	1838	speed for example), so if C<interval> is very small then timing stability
1584	will of course deteriorate. Libev itself tries to be exact to be about one	1839	will of course deteriorate. Libev itself tries to be exact to be about one
1585	millisecond (if the OS supports it and the machine is fast enough).	1840	millisecond (if the OS supports it and the machine is fast enough).
1586		1841
1587	=item * manual reschedule mode (at and interval ignored, reschedule_cb = callback)	1842	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
1588		1843
1589	In this mode the values for C<interval> and C<at> are both being	1844	In this mode the values for C<interval> and C<offset> are both being
1590	ignored. Instead, each time the periodic watcher gets scheduled, the	1845	ignored. Instead, each time the periodic watcher gets scheduled, the
1591	reschedule callback will be called with the watcher as first, and the	1846	reschedule callback will be called with the watcher as first, and the
1592	current time as second argument.	1847	current time as second argument.
1593		1848
1594	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	1849	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
1595	ever, or make ANY event loop modifications whatsoever>.	1850	or make ANY other event loop modifications whatsoever, unless explicitly
		1851	allowed by documentation here>.
1596		1852
1597	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop	1853	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
1598	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the	1854	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
1599	only event loop modification you are allowed to do).	1855	only event loop modification you are allowed to do).
1600		1856
…		…
1630	a different time than the last time it was called (e.g. in a crond like	1886	a different time than the last time it was called (e.g. in a crond like
1631	program when the crontabs have changed).	1887	program when the crontabs have changed).
1632		1888
1633	=item ev_tstamp ev_periodic_at (ev_periodic *)	1889	=item ev_tstamp ev_periodic_at (ev_periodic *)
1634		1890
1635	When active, returns the absolute time that the watcher is supposed to	1891	When active, returns the absolute time that the watcher is supposed
1636	trigger next.	1892	to trigger next. This is not the same as the C<offset> argument to
		1893	C<ev_periodic_set>, but indeed works even in interval and manual
		1894	rescheduling modes.
1637		1895
1638	=item ev_tstamp offset [read-write]	1896	=item ev_tstamp offset [read-write]
1639		1897
1640	When repeating, this contains the offset value, otherwise this is the	1898	When repeating, this contains the offset value, otherwise this is the
1641	absolute point in time (the C<at> value passed to C<ev_periodic_set>).	1899	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		1900	although libev might modify this value for better numerical stability).
1642		1901
1643	Can be modified any time, but changes only take effect when the periodic	1902	Can be modified any time, but changes only take effect when the periodic
1644	timer fires or C<ev_periodic_again> is being called.	1903	timer fires or C<ev_periodic_again> is being called.
1645		1904
1646	=item ev_tstamp interval [read-write]	1905	=item ev_tstamp interval [read-write]
…		…
1755	some child status changes (most typically when a child of yours dies or	2014	some child status changes (most typically when a child of yours dies or
1756	exits). It is permissible to install a child watcher I<after> the child	2015	exits). It is permissible to install a child watcher I<after> the child
1757	has been forked (which implies it might have already exited), as long	2016	has been forked (which implies it might have already exited), as long
1758	as the event loop isn't entered (or is continued from a watcher), i.e.,	2017	as the event loop isn't entered (or is continued from a watcher), i.e.,
1759	forking and then immediately registering a watcher for the child is fine,	2018	forking and then immediately registering a watcher for the child is fine,
1760	but forking and registering a watcher a few event loop iterations later is	2019	but forking and registering a watcher a few event loop iterations later or
1761	not.	2020	in the next callback invocation is not.
1762		2021
1763	Only the default event loop is capable of handling signals, and therefore	2022	Only the default event loop is capable of handling signals, and therefore
1764	you can only register child watchers in the default event loop.	2023	you can only register child watchers in the default event loop.
1765		2024
1766	=head3 Process Interaction	2025	=head3 Process Interaction
…		…
1852		2111
1853		2112
1854	=head2 C<ev_stat> - did the file attributes just change?	2113	=head2 C<ev_stat> - did the file attributes just change?
1855		2114
1856	This watches a file system path for attribute changes. That is, it calls	2115	This watches a file system path for attribute changes. That is, it calls
1857	C<stat> regularly (or when the OS says it changed) and sees if it changed	2116	C<stat> on that path in regular intervals (or when the OS says it changed)
1858	compared to the last time, invoking the callback if it did.	2117	and sees if it changed compared to the last time, invoking the callback if
		2118	it did.
1859		2119
1860	The path does not need to exist: changing from "path exists" to "path does	2120	The path does not need to exist: changing from "path exists" to "path does
1861	not exist" is a status change like any other. The condition "path does	2121	not exist" is a status change like any other. The condition "path does not
1862	not exist" is signified by the C<st_nlink> field being zero (which is	2122	exist" (or more correctly "path cannot be stat'ed") is signified by the
1863	otherwise always forced to be at least one) and all the other fields of	2123	C<st_nlink> field being zero (which is otherwise always forced to be at
1864	the stat buffer having unspecified contents.	2124	least one) and all the other fields of the stat buffer having unspecified
		2125	contents.
1865		2126
1866	The path I<should> be absolute and I<must not> end in a slash. If it is	2127	The path I<must not> end in a slash or contain special components such as
		2128	C<.> or C<..>. The path I<should> be absolute: If it is relative and
1867	relative and your working directory changes, the behaviour is undefined.	2129	your working directory changes, then the behaviour is undefined.
1868		2130
1869	Since there is no standard kernel interface to do this, the portable	2131	Since there is no portable change notification interface available, the
1870	implementation simply calls C<stat (2)> regularly on the path to see if	2132	portable implementation simply calls C<stat(2)> regularly on the path
1871	it changed somehow. You can specify a recommended polling interval for	2133	to see if it changed somehow. You can specify a recommended polling
1872	this case. If you specify a polling interval of C<0> (highly recommended!)	2134	interval for this case. If you specify a polling interval of C<0> (highly
1873	then a I<suitable, unspecified default> value will be used (which	2135	recommended!) then a I<suitable, unspecified default> value will be used
1874	you can expect to be around five seconds, although this might change	2136	(which you can expect to be around five seconds, although this might
1875	dynamically). Libev will also impose a minimum interval which is currently	2137	change dynamically). Libev will also impose a minimum interval which is
1876	around C<0.1>, but thats usually overkill.	2138	currently around C<0.1>, but that's usually overkill.
1877		2139
1878	This watcher type is not meant for massive numbers of stat watchers,	2140	This watcher type is not meant for massive numbers of stat watchers,
1879	as even with OS-supported change notifications, this can be	2141	as even with OS-supported change notifications, this can be
1880	resource-intensive.	2142	resource-intensive.
1881		2143
1882	At the time of this writing, the only OS-specific interface implemented	2144	At the time of this writing, the only OS-specific interface implemented
1883	is the Linux inotify interface (implementing kqueue support is left as	2145	is the Linux inotify interface (implementing kqueue support is left as an
1884	an exercise for the reader. Note, however, that the author sees no way	2146	exercise for the reader. Note, however, that the author sees no way of
1885	of implementing C<ev_stat> semantics with kqueue).	2147	implementing C<ev_stat> semantics with kqueue, except as a hint).
1886		2148
1887	=head3 ABI Issues (Largefile Support)	2149	=head3 ABI Issues (Largefile Support)
1888		2150
1889	Libev by default (unless the user overrides this) uses the default	2151	Libev by default (unless the user overrides this) uses the default
1890	compilation environment, which means that on systems with large file	2152	compilation environment, which means that on systems with large file
1891	support disabled by default, you get the 32 bit version of the stat	2153	support disabled by default, you get the 32 bit version of the stat
1892	structure. When using the library from programs that change the ABI to	2154	structure. When using the library from programs that change the ABI to
1893	use 64 bit file offsets the programs will fail. In that case you have to	2155	use 64 bit file offsets the programs will fail. In that case you have to
1894	compile libev with the same flags to get binary compatibility. This is	2156	compile libev with the same flags to get binary compatibility. This is
1895	obviously the case with any flags that change the ABI, but the problem is	2157	obviously the case with any flags that change the ABI, but the problem is
1896	most noticeably disabled with ev_stat and large file support.	2158	most noticeably displayed with ev_stat and large file support.
1897		2159
1898	The solution for this is to lobby your distribution maker to make large	2160	The solution for this is to lobby your distribution maker to make large
1899	file interfaces available by default (as e.g. FreeBSD does) and not	2161	file interfaces available by default (as e.g. FreeBSD does) and not
1900	optional. Libev cannot simply switch on large file support because it has	2162	optional. Libev cannot simply switch on large file support because it has
1901	to exchange stat structures with application programs compiled using the	2163	to exchange stat structures with application programs compiled using the
1902	default compilation environment.	2164	default compilation environment.
1903		2165
1904	=head3 Inotify and Kqueue	2166	=head3 Inotify and Kqueue
1905		2167
1906	When C<inotify (7)> support has been compiled into libev (generally	2168	When C<inotify (7)> support has been compiled into libev and present at
1907	only available with Linux 2.6.25 or above due to bugs in earlier	2169	runtime, it will be used to speed up change detection where possible. The
1908	implementations) and present at runtime, it will be used to speed up	2170	inotify descriptor will be created lazily when the first C<ev_stat>
1909	change detection where possible. The inotify descriptor will be created	2171	watcher is being started.
1910	lazily when the first C<ev_stat> watcher is being started.
1911		2172
1912	Inotify presence does not change the semantics of C<ev_stat> watchers	2173	Inotify presence does not change the semantics of C<ev_stat> watchers
1913	except that changes might be detected earlier, and in some cases, to avoid	2174	except that changes might be detected earlier, and in some cases, to avoid
1914	making regular C<stat> calls. Even in the presence of inotify support	2175	making regular C<stat> calls. Even in the presence of inotify support
1915	there are many cases where libev has to resort to regular C<stat> polling,	2176	there are many cases where libev has to resort to regular C<stat> polling,
1916	but as long as the path exists, libev usually gets away without polling.	2177	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2178	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2179	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2180	xfs are fully working) libev usually gets away without polling.
1917		2181
1918	There is no support for kqueue, as apparently it cannot be used to	2182	There is no support for kqueue, as apparently it cannot be used to
1919	implement this functionality, due to the requirement of having a file	2183	implement this functionality, due to the requirement of having a file
1920	descriptor open on the object at all times, and detecting renames, unlinks	2184	descriptor open on the object at all times, and detecting renames, unlinks
1921	etc. is difficult.	2185	etc. is difficult.
1922		2186
		2187	=head3 C<stat ()> is a synchronous operation
		2188
		2189	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2190	the process. The exception are C<ev_stat> watchers - those call C<stat
		2191	()>, which is a synchronous operation.
		2192
		2193	For local paths, this usually doesn't matter: unless the system is very
		2194	busy or the intervals between stat's are large, a stat call will be fast,
		2195	as the path data is usually in memory already (except when starting the
		2196	watcher).
		2197
		2198	For networked file systems, calling C<stat ()> can block an indefinite
		2199	time due to network issues, and even under good conditions, a stat call
		2200	often takes multiple milliseconds.
		2201
		2202	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2203	paths, although this is fully supported by libev.
		2204
1923	=head3 The special problem of stat time resolution	2205	=head3 The special problem of stat time resolution
1924		2206
1925	The C<stat ()> system call only supports full-second resolution portably, and	2207	The C<stat ()> system call only supports full-second resolution portably,
1926	even on systems where the resolution is higher, most file systems still	2208	and even on systems where the resolution is higher, most file systems
1927	only support whole seconds.	2209	still only support whole seconds.
1928		2210
1929	That means that, if the time is the only thing that changes, you can	2211	That means that, if the time is the only thing that changes, you can
1930	easily miss updates: on the first update, C<ev_stat> detects a change and	2212	easily miss updates: on the first update, C<ev_stat> detects a change and
1931	calls your callback, which does something. When there is another update	2213	calls your callback, which does something. When there is another update
1932	within the same second, C<ev_stat> will be unable to detect unless the	2214	within the same second, C<ev_stat> will be unable to detect unless the
…		…
2075		2357
2076	=head3 Watcher-Specific Functions and Data Members	2358	=head3 Watcher-Specific Functions and Data Members
2077		2359
2078	=over 4	2360	=over 4
2079		2361
2080	=item ev_idle_init (ev_signal *, callback)	2362	=item ev_idle_init (ev_idle *, callback)
2081		2363
2082	Initialises and configures the idle watcher - it has no parameters of any	2364	Initialises and configures the idle watcher - it has no parameters of any
2083	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2365	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
2084	believe me.	2366	believe me.
2085		2367
…		…
2098	// no longer anything immediate to do.	2380	// no longer anything immediate to do.
2099	}	2381	}
2100		2382
2101	ev_idle *idle_watcher = malloc (sizeof (ev_idle));	2383	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
2102	ev_idle_init (idle_watcher, idle_cb);	2384	ev_idle_init (idle_watcher, idle_cb);
2103	ev_idle_start (loop, idle_cb);	2385	ev_idle_start (loop, idle_watcher);
2104		2386
2105		2387
2106	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!	2388	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
2107		2389
2108	Prepare and check watchers are usually (but not always) used in pairs:	2390	Prepare and check watchers are usually (but not always) used in pairs:
…		…
2201	struct pollfd fds [nfd];	2483	struct pollfd fds [nfd];
2202	// actual code will need to loop here and realloc etc.	2484	// actual code will need to loop here and realloc etc.
2203	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));	2485	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
2204		2486
2205	/* the callback is illegal, but won't be called as we stop during check */	2487	/* the callback is illegal, but won't be called as we stop during check */
2206	ev_timer_init (&tw, 0, timeout * 1e-3);	2488	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
2207	ev_timer_start (loop, &tw);	2489	ev_timer_start (loop, &tw);
2208		2490
2209	// create one ev_io per pollfd	2491	// create one ev_io per pollfd
2210	for (int i = 0; i < nfd; ++i)	2492	for (int i = 0; i < nfd; ++i)
2211	{	2493	{
…		…
2324	some fds have to be watched and handled very quickly (with low latency),	2606	some fds have to be watched and handled very quickly (with low latency),
2325	and even priorities and idle watchers might have too much overhead. In	2607	and even priorities and idle watchers might have too much overhead. In
2326	this case you would put all the high priority stuff in one loop and all	2608	this case you would put all the high priority stuff in one loop and all
2327	the rest in a second one, and embed the second one in the first.	2609	the rest in a second one, and embed the second one in the first.
2328		2610
2329	As long as the watcher is active, the callback will be invoked every time	2611	As long as the watcher is active, the callback will be invoked every
2330	there might be events pending in the embedded loop. The callback must then	2612	time there might be events pending in the embedded loop. The callback
2331	call C<ev_embed_sweep (mainloop, watcher)> to make a single sweep and invoke	2613	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
2332	their callbacks (you could also start an idle watcher to give the embedded	2614	sweep and invoke their callbacks (the callback doesn't need to invoke the
2333	loop strictly lower priority for example). You can also set the callback	2615	C<ev_embed_sweep> function directly, it could also start an idle watcher
2334	to C<0>, in which case the embed watcher will automatically execute the	2616	to give the embedded loop strictly lower priority for example).
2335	embedded loop sweep.
2336		2617
2337	As long as the watcher is started it will automatically handle events. The	2618	You can also set the callback to C<0>, in which case the embed watcher
2338	callback will be invoked whenever some events have been handled. You can	2619	will automatically execute the embedded loop sweep whenever necessary.
2339	set the callback to C<0> to avoid having to specify one if you are not
2340	interested in that.
2341		2620
2342	Also, there have not currently been made special provisions for forking:	2621	Fork detection will be handled transparently while the C<ev_embed> watcher
2343	when you fork, you not only have to call C<ev_loop_fork> on both loops,	2622	is active, i.e., the embedded loop will automatically be forked when the
2344	but you will also have to stop and restart any C<ev_embed> watchers	2623	embedding loop forks. In other cases, the user is responsible for calling
2345	yourself - but you can use a fork watcher to handle this automatically,	2624	C<ev_loop_fork> on the embedded loop.
2346	and future versions of libev might do just that.
2347		2625
2348	Unfortunately, not all backends are embeddable: only the ones returned by	2626	Unfortunately, not all backends are embeddable: only the ones returned by
2349	C<ev_embeddable_backends> are, which, unfortunately, does not include any	2627	C<ev_embeddable_backends> are, which, unfortunately, does not include any
2350	portable one.	2628	portable one.
2351		2629
…		…
2445	event loop blocks next and before C<ev_check> watchers are being called,	2723	event loop blocks next and before C<ev_check> watchers are being called,
2446	and only in the child after the fork. If whoever good citizen calling	2724	and only in the child after the fork. If whoever good citizen calling
2447	C<ev_default_fork> cheats and calls it in the wrong process, the fork	2725	C<ev_default_fork> cheats and calls it in the wrong process, the fork
2448	handlers will be invoked, too, of course.	2726	handlers will be invoked, too, of course.
2449		2727
		2728	=head3 The special problem of life after fork - how is it possible?
		2729
		2730	Most uses of C<fork()> consist of forking, then some simple calls to ste
		2731	up/change the process environment, followed by a call to C<exec()>. This
		2732	sequence should be handled by libev without any problems.
		2733
		2734	This changes when the application actually wants to do event handling
		2735	in the child, or both parent in child, in effect "continuing" after the
		2736	fork.
		2737
		2738	The default mode of operation (for libev, with application help to detect
		2739	forks) is to duplicate all the state in the child, as would be expected
		2740	when I<either> the parent I<or> the child process continues.
		2741
		2742	When both processes want to continue using libev, then this is usually the
		2743	wrong result. In that case, usually one process (typically the parent) is
		2744	supposed to continue with all watchers in place as before, while the other
		2745	process typically wants to start fresh, i.e. without any active watchers.
		2746
		2747	The cleanest and most efficient way to achieve that with libev is to
		2748	simply create a new event loop, which of course will be "empty", and
		2749	use that for new watchers. This has the advantage of not touching more
		2750	memory than necessary, and thus avoiding the copy-on-write, and the
		2751	disadvantage of having to use multiple event loops (which do not support
		2752	signal watchers).
		2753
		2754	When this is not possible, or you want to use the default loop for
		2755	other reasons, then in the process that wants to start "fresh", call
		2756	C<ev_default_destroy ()> followed by C<ev_default_loop (...)>. Destroying
		2757	the default loop will "orphan" (not stop) all registered watchers, so you
		2758	have to be careful not to execute code that modifies those watchers. Note
		2759	also that in that case, you have to re-register any signal watchers.
		2760
2450	=head3 Watcher-Specific Functions and Data Members	2761	=head3 Watcher-Specific Functions and Data Members
2451		2762
2452	=over 4	2763	=over 4
2453		2764
2454	=item ev_fork_init (ev_signal *, callback)	2765	=item ev_fork_init (ev_signal *, callback)
…		…
2571	=over 4	2882	=over 4
2572		2883
2573	=item ev_async_init (ev_async *, callback)	2884	=item ev_async_init (ev_async *, callback)
2574		2885
2575	Initialises and configures the async watcher - it has no parameters of any	2886	Initialises and configures the async watcher - it has no parameters of any
2576	kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,	2887	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
2577	trust me.	2888	trust me.
2578		2889
2579	=item ev_async_send (loop, ev_async *)	2890	=item ev_async_send (loop, ev_async *)
2580		2891
2581	Sends/signals/activates the given C<ev_async> watcher, that is, feeds	2892	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2582	an C<EV_ASYNC> event on the watcher into the event loop. Unlike	2893	an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2583	C<ev_feed_event>, this call is safe to do from other threads, signal or	2894	C<ev_feed_event>, this call is safe to do from other threads, signal or
2584	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding	2895	similar contexts (see the discussion of C<EV_ATOMIC_T> in the embedding
2585	section below on what exactly this means).	2896	section below on what exactly this means).
2586		2897
		2898	Note that, as with other watchers in libev, multiple events might get
		2899	compressed into a single callback invocation (another way to look at this
		2900	is that C<ev_async> watchers are level-triggered, set on C<ev_async_send>,
		2901	reset when the event loop detects that).
		2902
2587	This call incurs the overhead of a system call only once per loop iteration,	2903	This call incurs the overhead of a system call only once per event loop
2588	so while the overhead might be noticeable, it doesn't apply to repeated	2904	iteration, so while the overhead might be noticeable, it doesn't apply to
2589	calls to C<ev_async_send>.	2905	repeated calls to C<ev_async_send> for the same event loop.
2590		2906
2591	=item bool = ev_async_pending (ev_async *)	2907	=item bool = ev_async_pending (ev_async *)
2592		2908
2593	Returns a non-zero value when C<ev_async_send> has been called on the	2909	Returns a non-zero value when C<ev_async_send> has been called on the
2594	watcher but the event has not yet been processed (or even noted) by the	2910	watcher but the event has not yet been processed (or even noted) by the
…		…
2597	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When	2913	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
2598	the loop iterates next and checks for the watcher to have become active,	2914	the loop iterates next and checks for the watcher to have become active,
2599	it will reset the flag again. C<ev_async_pending> can be used to very	2915	it will reset the flag again. C<ev_async_pending> can be used to very
2600	quickly check whether invoking the loop might be a good idea.	2916	quickly check whether invoking the loop might be a good idea.
2601		2917
2602	Not that this does I<not> check whether the watcher itself is pending, only	2918	Not that this does I<not> check whether the watcher itself is pending,
2603	whether it has been requested to make this watcher pending.	2919	only whether it has been requested to make this watcher pending: there
		2920	is a time window between the event loop checking and resetting the async
		2921	notification, and the callback being invoked.
2604		2922
2605	=back	2923	=back
2606		2924
2607		2925
2608	=head1 OTHER FUNCTIONS	2926	=head1 OTHER FUNCTIONS
…		…
2787		3105
2788	myclass obj;	3106	myclass obj;
2789	ev::io iow;	3107	ev::io iow;
2790	iow.set <myclass, &myclass::io_cb> (&obj);	3108	iow.set <myclass, &myclass::io_cb> (&obj);
2791		3109
		3110	=item w->set (object *)
		3111
		3112	This is an B<experimental> feature that might go away in a future version.
		3113
		3114	This is a variation of a method callback - leaving out the method to call
		3115	will default the method to C<operator ()>, which makes it possible to use
		3116	functor objects without having to manually specify the C<operator ()> all
		3117	the time. Incidentally, you can then also leave out the template argument
		3118	list.
		3119
		3120	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		3121	int revents)>.
		3122
		3123	See the method-C<set> above for more details.
		3124
		3125	Example: use a functor object as callback.
		3126
		3127	struct myfunctor
		3128	{
		3129	void operator() (ev::io &w, int revents)
		3130	{
		3131	...
		3132	}
		3133	}
		3134
		3135	myfunctor f;
		3136
		3137	ev::io w;
		3138	w.set (&f);
		3139
2792	=item w->set<function> (void *data = 0)	3140	=item w->set<function> (void *data = 0)
2793		3141
2794	Also sets a callback, but uses a static method or plain function as	3142	Also sets a callback, but uses a static method or plain function as
2795	callback. The optional C<data> argument will be stored in the watcher's	3143	callback. The optional C<data> argument will be stored in the watcher's
2796	C<data> member and is free for you to use.	3144	C<data> member and is free for you to use.
…		…
2882	L<http://software.schmorp.de/pkg/EV>.	3230	L<http://software.schmorp.de/pkg/EV>.
2883		3231
2884	=item Python	3232	=item Python
2885		3233
2886	Python bindings can be found at L<http://code.google.com/p/pyev/>. It	3234	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
2887	seems to be quite complete and well-documented. Note, however, that the	3235	seems to be quite complete and well-documented.
2888	patch they require for libev is outright dangerous as it breaks the ABI
2889	for everybody else, and therefore, should never be applied in an installed
2890	libev (if python requires an incompatible ABI then it needs to embed
2891	libev).
2892		3236
2893	=item Ruby	3237	=item Ruby
2894		3238
2895	Tony Arcieri has written a ruby extension that offers access to a subset	3239	Tony Arcieri has written a ruby extension that offers access to a subset
2896	of the libev API and adds file handle abstractions, asynchronous DNS and	3240	of the libev API and adds file handle abstractions, asynchronous DNS and
2897	more on top of it. It can be found via gem servers. Its homepage is at	3241	more on top of it. It can be found via gem servers. Its homepage is at
2898	L<http://rev.rubyforge.org/>.	3242	L<http://rev.rubyforge.org/>.
2899		3243
		3244	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		3245	makes rev work even on mingw.
		3246
		3247	=item Haskell
		3248
		3249	A haskell binding to libev is available at
		3250	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		3251
2900	=item D	3252	=item D
2901		3253
2902	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to	3254	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
2903	be found at L<http://proj.llucax.com.ar/wiki/evd>.	3255	be found at L<http://proj.llucax.com.ar/wiki/evd>.
		3256
		3257	=item Ocaml
		3258
		3259	Erkki Seppala has written Ocaml bindings for libev, to be found at
		3260	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
2904		3261
2905	=back	3262	=back
2906		3263
2907		3264
2908	=head1 MACRO MAGIC	3265	=head1 MACRO MAGIC
…		…
3009		3366
3010	#define EV_STANDALONE 1	3367	#define EV_STANDALONE 1
3011	#include "ev.h"	3368	#include "ev.h"
3012		3369
3013	Both header files and implementation files can be compiled with a C++	3370	Both header files and implementation files can be compiled with a C++
3014	compiler (at least, thats a stated goal, and breakage will be treated	3371	compiler (at least, that's a stated goal, and breakage will be treated
3015	as a bug).	3372	as a bug).
3016		3373
3017	You need the following files in your source tree, or in a directory	3374	You need the following files in your source tree, or in a directory
3018	in your include path (e.g. in libev/ when using -Ilibev):	3375	in your include path (e.g. in libev/ when using -Ilibev):
3019		3376
…		…
3075	keeps libev from including F<config.h>, and it also defines dummy	3432	keeps libev from including F<config.h>, and it also defines dummy
3076	implementations for some libevent functions (such as logging, which is not	3433	implementations for some libevent functions (such as logging, which is not
3077	supported). It will also not define any of the structs usually found in	3434	supported). It will also not define any of the structs usually found in
3078	F<event.h> that are not directly supported by the libev core alone.	3435	F<event.h> that are not directly supported by the libev core alone.
3079		3436
		3437	In stanbdalone mode, libev will still try to automatically deduce the
		3438	configuration, but has to be more conservative.
		3439
3080	=item EV_USE_MONOTONIC	3440	=item EV_USE_MONOTONIC
3081		3441
3082	If defined to be C<1>, libev will try to detect the availability of the	3442	If defined to be C<1>, libev will try to detect the availability of the
3083	monotonic clock option at both compile time and runtime. Otherwise no use	3443	monotonic clock option at both compile time and runtime. Otherwise no
3084	of the monotonic clock option will be attempted. If you enable this, you	3444	use of the monotonic clock option will be attempted. If you enable this,
3085	usually have to link against librt or something similar. Enabling it when	3445	you usually have to link against librt or something similar. Enabling it
3086	the functionality isn't available is safe, though, although you have	3446	when the functionality isn't available is safe, though, although you have
3087	to make sure you link against any libraries where the C<clock_gettime>	3447	to make sure you link against any libraries where the C<clock_gettime>
3088	function is hiding in (often F<-lrt>).	3448	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
3089		3449
3090	=item EV_USE_REALTIME	3450	=item EV_USE_REALTIME
3091		3451
3092	If defined to be C<1>, libev will try to detect the availability of the	3452	If defined to be C<1>, libev will try to detect the availability of the
3093	real-time clock option at compile time (and assume its availability at	3453	real-time clock option at compile time (and assume its availability
3094	runtime if successful). Otherwise no use of the real-time clock option will	3454	at runtime if successful). Otherwise no use of the real-time clock
3095	be attempted. This effectively replaces C<gettimeofday> by C<clock_get	3455	option will be attempted. This effectively replaces C<gettimeofday>
3096	(CLOCK_REALTIME, ...)> and will not normally affect correctness. See the	3456	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
3097	note about libraries in the description of C<EV_USE_MONOTONIC>, though.	3457	correctness. See the note about libraries in the description of
		3458	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		3459	C<EV_USE_CLOCK_SYSCALL>.
		3460
		3461	=item EV_USE_CLOCK_SYSCALL
		3462
		3463	If defined to be C<1>, libev will try to use a direct syscall instead
		3464	of calling the system-provided C<clock_gettime> function. This option
		3465	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		3466	unconditionally pulls in C<libpthread>, slowing down single-threaded
		3467	programs needlessly. Using a direct syscall is slightly slower (in
		3468	theory), because no optimised vdso implementation can be used, but avoids
		3469	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		3470	higher, as it simplifies linking (no need for C<-lrt>).
3098		3471
3099	=item EV_USE_NANOSLEEP	3472	=item EV_USE_NANOSLEEP
3100		3473
3101	If defined to be C<1>, libev will assume that C<nanosleep ()> is available	3474	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
3102	and will use it for delays. Otherwise it will use C<select ()>.	3475	and will use it for delays. Otherwise it will use C<select ()>.
…		…
3118		3491
3119	=item EV_SELECT_USE_FD_SET	3492	=item EV_SELECT_USE_FD_SET
3120		3493
3121	If defined to C<1>, then the select backend will use the system C<fd_set>	3494	If defined to C<1>, then the select backend will use the system C<fd_set>
3122	structure. This is useful if libev doesn't compile due to a missing	3495	structure. This is useful if libev doesn't compile due to a missing
3123	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout on	3496	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
3124	exotic systems. This usually limits the range of file descriptors to some	3497	on exotic systems. This usually limits the range of file descriptors to
3125	low limit such as 1024 or might have other limitations (winsocket only	3498	some low limit such as 1024 or might have other limitations (winsocket
3126	allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation, might	3499	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
3127	influence the size of the C<fd_set> used.	3500	configures the maximum size of the C<fd_set>.
3128		3501
3129	=item EV_SELECT_IS_WINSOCKET	3502	=item EV_SELECT_IS_WINSOCKET
3130		3503
3131	When defined to C<1>, the select backend will assume that	3504	When defined to C<1>, the select backend will assume that
3132	select/socket/connect etc. don't understand file descriptors but	3505	select/socket/connect etc. don't understand file descriptors but
…		…
3491	loop, as long as you don't confuse yourself). The only exception is that	3864	loop, as long as you don't confuse yourself). The only exception is that
3492	you must not do this from C<ev_periodic> reschedule callbacks.	3865	you must not do this from C<ev_periodic> reschedule callbacks.
3493		3866
3494	Care has been taken to ensure that libev does not keep local state inside	3867	Care has been taken to ensure that libev does not keep local state inside
3495	C<ev_loop>, and other calls do not usually allow for coroutine switches as	3868	C<ev_loop>, and other calls do not usually allow for coroutine switches as
3496	they do not clal any callbacks.	3869	they do not call any callbacks.
3497		3870
3498	=head2 COMPILER WARNINGS	3871	=head2 COMPILER WARNINGS
3499		3872
3500	Depending on your compiler and compiler settings, you might get no or a	3873	Depending on your compiler and compiler settings, you might get no or a
3501	lot of warnings when compiling libev code. Some people are apparently	3874	lot of warnings when compiling libev code. Some people are apparently
…		…
3535	==2274== definitely lost: 0 bytes in 0 blocks.	3908	==2274== definitely lost: 0 bytes in 0 blocks.
3536	==2274== possibly lost: 0 bytes in 0 blocks.	3909	==2274== possibly lost: 0 bytes in 0 blocks.
3537	==2274== still reachable: 256 bytes in 1 blocks.	3910	==2274== still reachable: 256 bytes in 1 blocks.
3538		3911
3539	Then there is no memory leak, just as memory accounted to global variables	3912	Then there is no memory leak, just as memory accounted to global variables
3540	is not a memleak - the memory is still being refernced, and didn't leak.	3913	is not a memleak - the memory is still being referenced, and didn't leak.
3541		3914
3542	Similarly, under some circumstances, valgrind might report kernel bugs	3915	Similarly, under some circumstances, valgrind might report kernel bugs
3543	as if it were a bug in libev (e.g. in realloc or in the poll backend,	3916	as if it were a bug in libev (e.g. in realloc or in the poll backend,
3544	although an acceptable workaround has been found here), or it might be	3917	although an acceptable workaround has been found here), or it might be
3545	confused.	3918	confused.
…		…
3574	way (note also that glib is the slowest event library known to man).	3947	way (note also that glib is the slowest event library known to man).
3575		3948
3576	There is no supported compilation method available on windows except	3949	There is no supported compilation method available on windows except
3577	embedding it into other applications.	3950	embedding it into other applications.
3578		3951
		3952	Sensible signal handling is officially unsupported by Microsoft - libev
		3953	tries its best, but under most conditions, signals will simply not work.
		3954
3579	Not a libev limitation but worth mentioning: windows apparently doesn't	3955	Not a libev limitation but worth mentioning: windows apparently doesn't
3580	accept large writes: instead of resulting in a partial write, windows will	3956	accept large writes: instead of resulting in a partial write, windows will
3581	either accept everything or return C<ENOBUFS> if the buffer is too large,	3957	either accept everything or return C<ENOBUFS> if the buffer is too large,
3582	so make sure you only write small amounts into your sockets (less than a	3958	so make sure you only write small amounts into your sockets (less than a
3583	megabyte seems safe, but this apparently depends on the amount of memory	3959	megabyte seems safe, but this apparently depends on the amount of memory
…		…
3587	the abysmal performance of winsockets, using a large number of sockets	3963	the abysmal performance of winsockets, using a large number of sockets
3588	is not recommended (and not reasonable). If your program needs to use	3964	is not recommended (and not reasonable). If your program needs to use
3589	more than a hundred or so sockets, then likely it needs to use a totally	3965	more than a hundred or so sockets, then likely it needs to use a totally
3590	different implementation for windows, as libev offers the POSIX readiness	3966	different implementation for windows, as libev offers the POSIX readiness
3591	notification model, which cannot be implemented efficiently on windows	3967	notification model, which cannot be implemented efficiently on windows
3592	(Microsoft monopoly games).	3968	(due to Microsoft monopoly games).
3593		3969
3594	A typical way to use libev under windows is to embed it (see the embedding	3970	A typical way to use libev under windows is to embed it (see the embedding
3595	section for details) and use the following F<evwrap.h> header file instead	3971	section for details) and use the following F<evwrap.h> header file instead
3596	of F<ev.h>:	3972	of F<ev.h>:
3597		3973
…		…
3633		4009
3634	Early versions of winsocket's select only supported waiting for a maximum	4010	Early versions of winsocket's select only supported waiting for a maximum
3635	of C<64> handles (probably owning to the fact that all windows kernels	4011	of C<64> handles (probably owning to the fact that all windows kernels
3636	can only wait for C<64> things at the same time internally; Microsoft	4012	can only wait for C<64> things at the same time internally; Microsoft
3637	recommends spawning a chain of threads and wait for 63 handles and the	4013	recommends spawning a chain of threads and wait for 63 handles and the
3638	previous thread in each. Great).	4014	previous thread in each. Sounds great!).
3639		4015
3640	Newer versions support more handles, but you need to define C<FD_SETSIZE>	4016	Newer versions support more handles, but you need to define C<FD_SETSIZE>
3641	to some high number (e.g. C<2048>) before compiling the winsocket select	4017	to some high number (e.g. C<2048>) before compiling the winsocket select
3642	call (which might be in libev or elsewhere, for example, perl does its own	4018	call (which might be in libev or elsewhere, for example, perl and many
3643	select emulation on windows).	4019	other interpreters do their own select emulation on windows).
3644		4020
3645	Another limit is the number of file descriptors in the Microsoft runtime	4021	Another limit is the number of file descriptors in the Microsoft runtime
3646	libraries, which by default is C<64> (there must be a hidden I<64> fetish	4022	libraries, which by default is C<64> (there must be a hidden I<64>
3647	or something like this inside Microsoft). You can increase this by calling	4023	fetish or something like this inside Microsoft). You can increase this
3648	C<_setmaxstdio>, which can increase this limit to C<2048> (another	4024	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
3649	arbitrary limit), but is broken in many versions of the Microsoft runtime	4025	(another arbitrary limit), but is broken in many versions of the Microsoft
3650	libraries.
3651
3652	This might get you to about C<512> or C<2048> sockets (depending on	4026	runtime libraries. This might get you to about C<512> or C<2048> sockets
3653	windows version and/or the phase of the moon). To get more, you need to	4027	(depending on windows version and/or the phase of the moon). To get more,
3654	wrap all I/O functions and provide your own fd management, but the cost of	4028	you need to wrap all I/O functions and provide your own fd management, but
3655	calling select (O(n²)) will likely make this unworkable.	4029	the cost of calling select (O(n²)) will likely make this unworkable.
3656		4030
3657	=back	4031	=back
3658		4032
3659	=head2 PORTABILITY REQUIREMENTS	4033	=head2 PORTABILITY REQUIREMENTS
3660		4034
…		…
3703	=item C<double> must hold a time value in seconds with enough accuracy	4077	=item C<double> must hold a time value in seconds with enough accuracy
3704		4078
3705	The type C<double> is used to represent timestamps. It is required to	4079	The type C<double> is used to represent timestamps. It is required to
3706	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	4080	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
3707	enough for at least into the year 4000. This requirement is fulfilled by	4081	enough for at least into the year 4000. This requirement is fulfilled by
3708	implementations implementing IEEE 754 (basically all existing ones).	4082	implementations implementing IEEE 754, which is basically all existing
		4083	ones. With IEEE 754 doubles, you get microsecond accuracy until at least
		4084	2200.
3709		4085
3710	=back	4086	=back
3711		4087
3712	If you know of other additional requirements drop me a note.	4088	If you know of other additional requirements drop me a note.
3713		4089
…		…
3781	involves iterating over all running async watchers or all signal numbers.	4157	involves iterating over all running async watchers or all signal numbers.
3782		4158
3783	=back	4159	=back
3784		4160
3785		4161
		4162	=head1 GLOSSARY
		4163
		4164	=over 4
		4165
		4166	=item active
		4167
		4168	A watcher is active as long as it has been started (has been attached to
		4169	an event loop) but not yet stopped (disassociated from the event loop).
		4170
		4171	=item application
		4172
		4173	In this document, an application is whatever is using libev.
		4174
		4175	=item callback
		4176
		4177	The address of a function that is called when some event has been
		4178	detected. Callbacks are being passed the event loop, the watcher that
		4179	received the event, and the actual event bitset.
		4180
		4181	=item callback invocation
		4182
		4183	The act of calling the callback associated with a watcher.
		4184
		4185	=item event
		4186
		4187	A change of state of some external event, such as data now being available
		4188	for reading on a file descriptor, time having passed or simply not having
		4189	any other events happening anymore.
		4190
		4191	In libev, events are represented as single bits (such as C<EV_READ> or
		4192	C<EV_TIMEOUT>).
		4193
		4194	=item event library
		4195
		4196	A software package implementing an event model and loop.
		4197
		4198	=item event loop
		4199
		4200	An entity that handles and processes external events and converts them
		4201	into callback invocations.
		4202
		4203	=item event model
		4204
		4205	The model used to describe how an event loop handles and processes
		4206	watchers and events.
		4207
		4208	=item pending
		4209
		4210	A watcher is pending as soon as the corresponding event has been detected,
		4211	and stops being pending as soon as the watcher will be invoked or its
		4212	pending status is explicitly cleared by the application.
		4213
		4214	A watcher can be pending, but not active. Stopping a watcher also clears
		4215	its pending status.
		4216
		4217	=item real time
		4218
		4219	The physical time that is observed. It is apparently strictly monotonic :)
		4220
		4221	=item wall-clock time
		4222
		4223	The time and date as shown on clocks. Unlike real time, it can actually
		4224	be wrong and jump forwards and backwards, e.g. when the you adjust your
		4225	clock.
		4226
		4227	=item watcher
		4228
		4229	A data structure that describes interest in certain events. Watchers need
		4230	to be started (attached to an event loop) before they can receive events.
		4231
		4232	=item watcher invocation
		4233
		4234	The act of calling the callback associated with a watcher.
		4235
		4236	=back
		4237
3786	=head1 AUTHOR	4238	=head1 AUTHOR
3787		4239
3788	Marc Lehmann <libev@schmorp.de>.	4240	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson.
3789		4241

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines